ChatGPT, Claude, Gemini & Deepseek all agreed "Engaging four AI models in a debate to reach consensus is estimated to improve reasoning capabilities by approximately 20-50%". They agreed through the product i'm announcing in this post: https://getconsensus.ai/ that according to them, makes AI reasoning up to 50% better. Here's how I ended up creating it: Been working for months on 2 major products. To tackle some complex technical problems, I was constantly prompting and refining responses from several AI models—ChatGPT, Claude, Gemini, Grok, etc. Individually, each model provided answers that were ~95% accurate, but each made different errors. However, chatting with the models or combining their responses usually produced near-perfect answers. Then I thought: What if I can just get all these AI models together, get them to each give their own response, debate amongst themselves, and collaboratively refine the answer until they reached a consensus? The result is apparently ~20-50% better reasoning. Originally planned as an internal tool, but I thought it was worth another couple of days to turn it into getconsensus.ai. I recommend using it for complex coding problems like smart contract development. It also solves easy questions where individual models fail like the screenshots shared. Full disclosure: this is still an experimental version, and I am currently trying to get higher rate limits from APIs. New sign ups get 5 free debates or even buy credits, if needed use your own api keys to use it for free forever and bypass rate limit issues. If you'd like to wait for a refined version, I'll announce beta version here soon.

GetConsensus.AI

Ask a complex question while AI models debate the answer

Ask a complex question while AI models debate the answer